Neuro sliding mode control of robotic manipulators

This paper considers neuro-adaptive control of robotic manipulators based on sliding mode technique. Since robotic manipulators are inherently nonlinear systems, neural network (NN) based controllers have been proposed as a feasible technique to achieve consistent performance, owing to the adaptive nonlinear learning capabilities of NNs. Chebyshev neural network (CNN) is employed to approximate the unknown nonlinearities associated with robotic manipulators. The design basically consists of a neural controller which implements the equivalent control law associated with sliding mode technique for the manipulator with unknown dynamics and an adaptive switching gain which provides robustness and compensates for the neural approximation errors and the disturbances in the system. A single layer CNN is used to realize the proposed controller whose weights are updated online and the Lyapunov stability theory is used to prove the ultimate uniform boundedness of the tracking error. Simulation results illustrate the effectiveness of the proposed method to achieve desired performance in the presence of bounded disturbances.